FAK-mediated extracellular signals are essential for interkinetic nuclear migration and planar divisions in the neuroepithelium

During the morphogenesis of the neural tube, neural progenitors continuously divide, and post mitotic neurons then migrate to form complex neural networks. It is widely conserved among vertebrates that this mitotic process takes place near the lumen, the apical side of the neural tube (Gotz and Huttner, 2005). When viewed at the cellular level, this apically localized mitosis is the region of output for the cell-cycle-dependent oscillation of the nuclei within the neuroepithelium. Mitotic (M phase) nuclei are located in close proximity to the apical surfaces, whereas nuclei undergoing DNA synthesis (S phase) are displaced more basally. This characteristic movement of nuclei is known as interkinetic nuclear migration (INM) (Fujita, 1964; Sauer, 1935) and is thought to be important for maintaining a neural progenitor pool during neurogenesis (Baye and Link, 2007; Del Bene et al., 2008; Murciano et al., 2002; Xie et al., 2007). Furthermore, planar cell divisions near the apical surface are thought to be important to maintain the progenitor pool (Gotz and Huttner, 2005). Indeed, recent works suggest that most divisions during both proliferative and neurogenic stages are planar (that is, parallel to the apical surface), whereas vertical or oblique divisions are rarely seen, and daughter cells that inherit both the apical and basal components are suggested to remain in the progenitor pool (Konno et al., 2008). Although INM and planar cell divisions have been regarded as hallmarks of vertebrate neural progenitors, their cellular and molecular mechanisms are just beginning to be elucidated. Previous studies have suggested the involvement of several intracellular components such as the actin and microtubule cytoskeletons, molecules involved in adherens junctions, polarity complex and the centrosome (Cappello et al., 2006; Imai et al., 2006; Lien et al., 2006; Murciano et al., 2002; Siller and Doe, 2009; Xie et al., 2007). However, it remains unclear whether extracellular molecules, such as components of the extracellular matrix (ECM), have any role in INM and mitotic orientation of neural progenitors, and if so, how they interact with the intracellular machinery that drives this process. Indeed, the ECM has long been shown to participate in neurogenesis (see below), but there are no reports of vertebrate ECM mutants with defective INM and mitotic orientation. In this context, the medaka mutant tacobo (tab) used in the present study provides a unique framework. The medaka fish, Oryzias latipes, is an emerging vertebrate model and now has a high quality draft genome and a number of unique mutants (Furutani-Seiki et al., 2004; Kasahara et al., 2007; Wittbrodt et al., 2002). Our recent mutagenesis screen in medaka identified tab as a mutant showing pleiotropic phenotypes including impaired INM in the neuroepithelium. Here, we report that tab encodes laminin γ1 (Lamc1) and that a mutation in this gene causes the impaired INM and mitotic orientation.

Laminins are heterotrimeric glycoprotein complexes consisting of individual α-, β- and γ-chain subunits, and are major structural components of basement membranes (BMs). To date, 15 mammalian laminin complexes have been described comprising five α-chains, three β-chains and three γ-chains. Among these different chain types, laminin γ1 is the most commonly found, contributing to ten laminins including laminin 1, which is essential for early embryogenesis (Miner and Yurchenco, 2004). In addition to their structural roles, the laminins act as signaling molecules through their cell surface receptors, which include integrins (Yurchenco et al., 2004). A variety of laminin functions have been elucidated through genetic studies in humans, mice and zebrafish, and include cell migration, proliferation, differentiation and cell polarity (Miner and Yurchenco, 2004; O'Brien et al., 2001). As an example, recent work in both mouse and human has revealed a role of laminin 5 (α3β3γ2) in the epidermis; mutations in one of the chains cause defective skin morphogenesis in mouse or Herlitz junctional epidermolysis bullosa in human (Aberdam et al., 1994; Ryan et al., 1999). Furthermore, studies of zebrafish laminin mutants have highlighted the broader role of these molecules in organogenesis of such structures as the notochord and blood vessels (Parsons et al., 2002; Pollard et al., 2006).

In the case of neurogenesis, the proper functioning of the laminins was found to be a prerequisite for the formation of the cortical BMs, defects in which cause impaired cortical histogenesis whereby cortical neurons protrude into the subarachnoid (Halfter et al., 2002). Genetic analyses of laminins is sometimes hampered by their requirement in early developmental processes and their functional redundancy; for example, a lack of laminin α1, β1 and γ1 in mice leads to early post implantation lethality (Alpy et al., 2005; Miner et al., 2004; Smyth et al., 1999). Moreover, compensation of one chain in the absence of another has been frequently reported (Miner et al., 2004; Yang et al., 2005). As a likely result of these difficulties, the role of laminins in early neurogenesis, particularly in INM and mitotic orientation, has not yet been fully addressed (Halfter et al., 2002; Haubst et al., 2006).

In our current study, we use medaka wild-type and tab mutant embryos for analysis of neurogenesis. We focus on the early stage neural tube in which mitotically active neuroepithelial cells exhibit typical INM: they maintain cytoplasmic connections to both the apical (ventricular) and basal surfaces, and their nuclei migrate over the entire thickness of the neuroepithelium. Taking advantage of the transparency of the medaka fish embryos and simple structure of their early neural tubes, a live-imaging technique was successfully applied to record the behavior of individual nuclei in both the wild-type and tab neuroepithelium. Our genetic and imaging analyses clearly show that INM and planar cell divisions are dependent upon FAK-mediated signals downstream of laminins. Furthermore, we reveal a genetic interaction between FAK-mediated signals and the dynein system for the regulation of INM and mitotic orientation. To our knowledge, this is the first report to genetically demonstrate the important role of ECM molecules in these processes, linking extracellular signals with the intracellular machinery.

The tacobo (tab) medaka mutant was recessive and isolated in our screening of ENU-induced mutant fish with disrupted axis elongation and neuronal development (see below) (Ishikawa, 1996). We undertook a positional cloning approach to identify the tab gene. Mapping analysis with EST markers indicated that tab was located close to the Ola2909h marker (4/1504 recombinants/meiosis) on LG17. Chromosome walking using BAC clones was then carried out to narrow down the responsible region. We found that the gene encoding Lamc1 mapped close to the responsible region and we therefore tested whether tab encodes Lamc1. A comparison of genomic laminin γ1 (lamc1) sequences between the wild-type (wt) and tab mutant revealed a T to A substitution in the presumptive splice-donor site (the 20th intron) in the mutant (Fig. 1A). This mutation produces three kinds of abnormal splicing products which were all detected in tab embryos by RT-PCR, in addition to two other unexpected fragments (Fig. 1B). The aberrant transcript gave rise to a presumptive truncated form of Lamc1 that lacks most of the coiled-coil domain within the C-terminus (Fig. 1B). Because the coiled-coil domain in the laminin chains is required for the formation of heterotrimers, the deduced truncated protein was expected to be nonfunctional (Beck et al., 1993). In addition, lamc1 is expressed almost ubiquitously during neurulation and neurogenesis with stronger expression in the midbrain-hindbrain boundary region, which is largely reduced in the tab mutant (data not shown). Furthermore, injection of two individual MOs specific for lamc1 (designed to target the first methionine and the splicing donor site that is affected in tab) phenocopied the tab phenotype (Fig. 1C). We thus concluded that the tab locus encodes the medaka laminin γ1 chain.

Tab is a recessive lethal mutant showing pleiotropic morphological defects such as small eye, a shortened body axis, an enlarged midbrain-hindbrain boundary, defective notochord vacuolation, loss of jaw and malformed pectoral fins (Fig. 1C,D). Blood vessel formation is also severely affected in the tab mutant as seen in the zebrafish lamc1 mutant sleepy (slp) (Fig. 1C). Furthermore, histological analysis revealed that tab develops defects in retinal lamination (Fig. 1D). These phenotypes are again reminiscent of the zebrafish laminin mutants, bashful (bal) (lama1), grumpy (gup) (lamb1) and slp (lamc1) (Parsons et al., 2002; Pollard et al., 2006), but the overall phenotype of tab appears to be more severe than any of the zebrafish single laminin mutants.

Since early neural tube patterning appears normal in the tab mutant as indicated by the normal expression of regional markers (wnt1, shh and krox20; supplementary material Fig. S1), the aberrant morphogenesis that manifests in the neural tube could therefore be the result of abnormal cell proliferation at earlier stages. We thus examined the frequency and distribution pattern of neuroepithelial cells in late G2-M phase at the 10-somite stage using phosphorylated histone H3 (PH3) as a marker. At this stage, the gross morphology of the neural tube still appears normal in the tab mutant. We first calculated the mitotic index (the ratio of PH3-positive nuclei to total nuclei in a defined area) and found no significant difference between the wild-type and tab neural tubes (Fig. 2B). Also, the cell-cycle length was found to be unaffected in the tab neural tube by the BrdU pulse-labeling analysis (Fig. 2B, supplementary material Fig. S2). Interestingly, a significant difference was observed in the positioning of the PH3-positive nuclei between wild type and tab. In wild-type medaka embryos, the labeled cells were mostly localized near the apical side of the neural tube (Fig. 2A). In the tab mutant, however, these cells were broadly dispersed along the apico-basal axis (Fig. 2A, arrowheads). The basal displacement of mitotic nuclei was observed along the entirety of the tab neural tube. This tendency was confirmed further by quantitative analysis (Fig. 2C): the relative distance of the labeled nuclei from the apical surface was found to have significantly increased in tab.

In the analysis below, we focused on this basal displacement of mitosis in the hindbrain region, because this phenotype has not been reported in laminin-related mutants in other vertebrate models, and because in the hindbrain region of tab, the phenotype is at its most pronounced and the imaging of nuclear migration can be readily performed.

To explore the role of lamc1 in the positioning of mitotic nuclei in the neuroepithelium, we first examined the laminin deposition around the neural tube using an antibody raised against laminin 1. Laminin 1 is a heterotrimer of α1, β1, γ1 laminin chains, and is the major heterotrimer of the laminin γ1 subunit. As reported previously in zebrafish embryos using the same antibody we used in this study (Parsons et al., 2002), strong laminin 1 immunoreactivity was observed around the basal surface of medaka wild-type neural tubes throughout the segmentation stages, whereas in tab, this immunoreactivity almost disappeared (Fig. 2D, left panels). This suggests that the loss of the laminin γ1 chain causes a severe reduction in laminin 1 accumulation on the basal side of the neural epithelium. This reduction in immunoreactivity occurs in various organs, including blood vessels (data not shown). As laminins are essential for basement membrane (BM) assembly (Miner and Yurchenco, 2004), we examined the ultrastructure of the BM beneath the neural tube. In wild-type embryos, the basal surface of the neural tube was covered by a continuous BM sheet. By contrast, in tab, thinner and disorganized BMs were widely observed (Fig. 2D, middle panels). We also found that the basal surface of the tab neural tube was not as smooth as that in the wild type. Disorganized BMs were also found in the notochord sheath of the zebrafish sly mutant and retina of Lamc1-knockout mice (Halfter et al., 2002; Parsons et al., 2002). By contrast, fibronectin, which was reported to be localized to the basal surface of the neural tube (Mayer et al., 1981), appeared to be normally distributed around the tab neural tube (Fig. 2D, right panels).

In our present study, we focused on integrins because laminin signals are, to a large extent, mediated by these molecules and laminin-integrin interactions have been implicated in neuronal cell migration and cell shape change (Colognato et al., 2005; Georges-Labouesse et al., 1998). Among the downstream targets of laminin-integrin signaling, the protein tyrosine kinase FAK is known to have a prominent role in neurogenesis (Beggs et al., 2003; Parsons, 2003; Xie et al., 2003). Since the phosphorylation of tyrosine on position 397 (Y397) is a hallmark of FAK activation, we investigated both its phosphorylation state and cellular distribution in the neural tube using an anti-phosphorylated FAK (pFAK) antibody (see supplementary material Fig. S3 for antibody specificity).

In wild-type embryos, phosphorylated FAK was accumulated at both the basal and apical side of the neuroepithelium. In tab embryos, however, the accumulation of pFAK is greatly downregulated on both sides of the neuroepithelium (Fig. 2E), suggesting that the activation of FAK mostly depends on laminin signals in the neural tube. The downregulation of FAK was also founded in the retina of zebrafish lama1 mutant bal (Semina et al., 2006). It is also worth noting that FAK was found to be normally activated at the somite boundary in tab, and consistently, no somite segmentation phenotype could be observed in tab embryos. This suggests that FAK is activated by various signaling pathways, such as fibronectin, depending on contexts and tissues (Koshida et al., 2005).

We next examined by morpholino (MO) knockdown to what degree FAK-mediated signaling accounts for neural tube phenotypes in tab. Fak-MO-injected embryos exhibited mislocalized mitosis along the entire neural tube as well as a shortened body axis and dysplastic eyes, a phenotype identical to that of tab (Fig. 2F). These results indicate that FAK-mediated signals indeed play a major role in lamc1-dependent morphogenesis, including the positioning of mitotic nuclei in the neuroepithelium.

To address the cellular mechanism underlying mislocalized mitosis in the tab mutant medaka, we tested whether the interkinetic nuclear migration (INM), a characteristic feature of neuroepithelial cells, was affected in tab neural tubes. For this analysis, we used confocal time-lapse imaging of neuroepithelial cells labeled with EGFP at around the 9- to 10-somite stages. In wild-type neural tubes, the typical nuclear migration associated with INM was seen, i.e. most nuclei had migrated from the basal side and divided when they reach the apical region (Fig. 3A, and supplementary material Movie 1). However, in tab mutant neural tubes, many nuclei exhibited abnormal and variable patterns of migration. A typical abnormal behavior in tab nuclei was slower apical migration and entry into mitosis prior to reaching the apical region (Fig. 3A, and supplementary material Movie 2). Superimposed traces demonstrate impaired nuclear migration around the time of mitosis in tab. Some nuclei showed normal migration, but in the most extreme cases, they migrated in an opposite direction (i.e. basally), followed by mitosis in the basal region (Fig. 3B). Most nuclei showed intermediate phenotypes, leading to mislocalized mitosis. Furthermore, we retrospectively measured the basal-to-apical migration velocity of the two populations of tab nuclei, one showing normal apical mitosis and the other showing basally displaced mitosis. In both cases, the migration velocity of tab nuclei was found to significantly decrease (Fig. 3C). These results indicate that abnormal INM in tab leads to the increased number of mislocalized mitoses.

We speculated that the reduced nuclear migration activity in the tab mutant could be the result of disrupted cell polarity and/or cell-shape change. Previous reports have demonstrated that the apico-basal polarity is a prerequisite for correct mitotic nuclei positioning, which possibly affects nuclear migration and spindle orientation (Horne-Badovinac et al., 2001; Imai et al., 2006), and that laminins have an essential role in the establishment of this polarity (Li et al., 2002; O'Brien et al., 2001). However, our analysis of a series of apical markers (aPKCζ, ZO-1, γ-tubulin) showed that they were correctly localized in the mutant neuroepithelium (Fig. 4A).

We next assessed the morphology of neuroepithelial cells, which have apical and basal cellular processes extending to the basal and apical surfaces (ventricular and pial surfaces), a typical characteristic of neuroepithelial cells. For this analysis, embryos were injected at the 8- to 16-cell stage with egfp mRNA to label cells in a mosaic manner. In both wild-type and tab neural tubes, almost all cells displayed cellular processes extending to the apical and basal surfaces of the neural tube (wt, 73/74, n=5; tab, 72/73, n=5) (Fig. 4B, left panel). The detached processes observed in the wild type and tab are both basal processes of newly born daughter cells extending their processes toward basal surface of the neural tube. The attachment was also confirmed with the fixed embryos injected with egfp mRNA and stained with phalloidin (Fig. 4B, right panels). The attachment of cells to apical and basal surface of the neural tube was also maintained in fak-MO-injected embryos (supplementary material Fig. S4; 70/70, n=5).

Taken together, both cell polarity and morphology of neuroepithelial cells are not affected in tab. Recent works suggested that several molecules such as integrin-linked kinase (ILK) are also important downstream molecules of integrin signaling and are essential for cell polarization and cell morphology (Legate et al., 2006; Sakai et al., 2003). We speculate that ILK could compensate for the disrupted BMs. The results thus do not support the idea that cell polarity and/or cell shape defects are the primary cause of the basally displaced mitotic events in tab neural tubes, suggesting that the intracellular machinery directing nuclear migration is impaired in these mutant cells.

Recent works show that integrin signals regulate spindle orientation in both polarized and non-polarized cells (Lechler and Fuchs, 2005; Toyoshima and Nishida, 2007). We thus examined whether cleavage-plane orientation is affected in the tab neural tube using time-lapse imaging (Fig. 5A). In wild-type neural tubes, most mitotic cells in the apical surface show vertical cleavage (planar division), i.e. the cleavage plane is vertical to the apical surface (79%, 57/72). By contrast, in tab, the cleavage plane becomes nearly randomized: 45% (34/75) for vertical, 19% (14/75) for oblique and 36% (27/75) for horizontal. In mislocalized mitosis, the cleavage plane was horizontal, irrespective of the genotype (wt, 2/2; tab, 31/31). The cleavage plane also appeared to be randomized when mitotic orientation was examined in fixed sections (Fig. 5A). The same tendency was obtained in fak-MO-injected embryos (Fig. 5A). These data indicate that FAK-mediated signaling is essential for planar division in the apical surface of the neuroepithelium.

To assess the impact of the above mentioned defects on neurogenesis, we first measured the mitotic index (percentage of phosphorylated histone-3-positive cells) in the neural tube at the 15-somite stage, the stage corresponding to the neurogenic phase. This index was significantly decreased in tab neural tubes compared with that in the wild type (Fig. 5B), suggesting that there is a reduced proportion of M-phase progenitors at this stage in tab. To determine whether this possible reduction in M-phase progenitors is accompanied by the increased number of postmitotic neurons, we stained differentiated neurons with Hu C/D, a pan neuronal maker at the 10-somite (proliferative phase) and 15-somite stages. Although no significant difference was observed at the 10-somite stage, a significant increase in Hu-positive cells was observed at the 15-somite stage (Fig. 5C and supplementary material Fig. S5), suggesting that accelerated neurogenesis is occurring in the tab neural tube during the neurogenic phase.

We also found some Hu-positive cells located outside the neural tubes of tab and fak morphants (i.e. ectopia), which was never observed in the wild type (Fig. 5C and supplementary material Fig. S6, arrowheads). Similar ectopias were also reported in the cortices of Lamc1-knockout mice, in which neurons were suggested to migrate outward from the neural tube through a gap in BMs of the neural tube (Halfter et al., 2002).

To test cell autonomy of the FAK function, we performed mosaic analysis using fak morphants. We used fak morphants instead of tab mutants, because the tab mutation was expected to function non-cell-autonomously. Compared with the control-MO-injected cells and host wild-type cells, fak-MO-cells, when transplanted, specifically exhibited a significant increase in mislocalized mitotic nuclei (Fig. 5D, middle panels, and supplementary material Fig. S7), planar cell division (Fig. 5D, right panels, and supplementary material Fig. S7) and accelerated neurogenesis (Fig. 5E, supplementary material Fig. S7). These results indicate the cell-autonomous function of FAK, and also suggest the correlation between defects in mitosis at the earlier stage and accelerated neurogenesis at the later stage.

Recent studies have implicated dynein-dynactin as a mediator for nuclear migration in retinal neuroepithelium, nuclear positioning in photoreceptor cells (Del Bene et al., 2008; Tsujikawa et al., 2007) and spindle orientation in various cell types (Siller and Doe, 2009; Toyoshima et al., 2007), highlighting the importance of the dynein system in these processes. Dynactin is a large multi-subunit activator of dynein, which mediates the interaction of its cargo and enhances dynein processivity (Gross, 2003; King and Schroer, 2000). These led us to hypothesize that the activity of dynein could be impaired in the neural tube of the tab mutant, leading to defects in INM and planar divisions. To test this hypothesis, we attempted to decrease the activity of the dynactin complex in medaka embryos. In several experimental systems, the overexpression of dynamitin (p50, Dynactin2), a dynactin component (King and Schroer, 2000), was found to cause a dissociation of the dynactin complex and consequently to block its function (Burkhardt et al., 1997). Taking advantage of this, we overexpressed dynamitin in medaka embryos by injecting medaka dynamitin mRNA at the one-cell stage. Overexpression of dynamitin produced randomized orientation of apical cell divisions (Fig. 6A) and basally localized mitosis at the 10-somite stage (Fig. 6A). The mislocalized mitosis was reminiscent of the phenotype of mikre oko (mok), a zebrafish dynactin mutant in which INM in the retina is defective (Del Bene et al., 2008). Similar to mok, dynamitin-injected embryos displayed a severe defect in eye morphogenesis at 4 d.p.f., confirming the block in dynactin function by dynamitin overexpression in medaka embryos (Fig. 6A). These results indicate that the dynein-dynactin complex indeed regulates INM and mitotic orientation in the neural tube.

As the inhibition of laminin-FAK and dynein-dynactin activities produced similar phenotypes, we investigated the possible interplay between these two signals by performing a genetic interaction study. dynamitin mRNA and fak MO were titrated to suboptimal doses (dynamitin mRNA, 50 ng/μl; fak MO, 100 μM) and co-injected into medaka embryos to inhibit both pathways. As shown in Fig. 6B, single injections of a suboptimal dose of either mRNA or MO caused no significant defects. However, co-injection caused significantly greater effects on the positioning of mitotic nuclei and mitotic orientation in the neural tube (Fig. 6B). Together, these results suggest that the dynein-microtubule system and FAK-mediated signaling cooperate for INM and planar divisions in the neural tube.

Finally, we examined the distribution of dynactin p150 by staining wild-type neural tubes with the anti-dynactin p150 antibody and found that dynactin is enriched in both the apical and basal regions of the neural tube, in addition to the low levels of uniform distribution in the cytoplasm. In the neural tubes of fak morphants, the dynactin signal appeared to be evenly reduced compared with the control-MO-injected embryos (supplementary material Figs S3, S8A). However, the distribution of phosphorylated FAK was not affected by dynamitin overexpression (supplementary material Fig. S8C). Together, these results strengthen the notion that the dynein-microtubule system works in the FAK pathway in the neural tube.

INM has been the focus of numerous studies of neurogenesis since it was first reported in 1930s (Del Bene et al., 2008; Murciano et al., 2002; Sauer, 1935; Xie et al., 2007). The importance of cleavage orientation of neural progenitors in maintaining self-renewability has also been recognized for many years (Gotz and Huttner, 2005; Konno et al., 2008). We have addressed the role of extracellular signals in these characteristic features of neurogenesis using the lamc1 medaka mutant tab. We demonstrate for the first time that extracellular signals mediated by FAK regulate the activity of nuclear migration and the mitotic orientation in the vertebrate neuroepithelium. Our data also suggest that the dynein motor system functions in these two processes, possibly through interaction with FAK signaling. The tab mutant thus reveals the novel role of extracellular signals during vertebrate early neurogenesis (Fig. 6C).

The tab locus was found to encode the medaka laminin γ1 chain, and a mutated tab locus produces abnormally spliced transcripts that are translated into truncated forms of laminin γ1 chain which lack a coiled-coil domain. This domain is essential for heterotrimer formation (Beck et al., 1993). Although RT-PCR detected two other transcripts of unexpected size, tab is likely to be a null allele because of the severe loss of laminin 1 reactivity and the successful phenocopy by MO injection. The phenotype of the tab mutants largely resembles that of the zebrafish laminin mutants, bal, gup and sly (Parsons et al., 2002; Pollard et al., 2006; Semina et al., 2006). However, defects in INM and planar divisions are found only in tab embryos. This might reflect a difference in the redundancy of the laminin isoforms and/or in the degree of the maternal contribution of laminins between the two fish systems. Indeed, lamc1 is maternally provided as RNA and protein both in zebrafish and medaka (data not shown), and this contribution might alleviate the neural tube phenotype more in fast-developing zebrafish than in the medaka. In spite of this difference, however, tab and sly both exhibit an abnormal MHB morphology at relatively later stages. We consider this not to be a direct consequence of defective INM but rather a failure in laminin-dependent basal constriction, as previously reported (Gutzman et al., 2008).

As in zebrafish sly mutants, we observed a large reduction in laminin 1 immunoreactivity in the tab medaka, most notably around the mutant neural tubes, which exhibit basally mislocalized mitosis and a randomized cleavage plane in the apical region. The simplest interpretation of this finding is that the reduction in laminin accumulation directly induces the observed INM and mitotic defects in the neuroepithelium. However, the tab situation might not be so simple. The BM is a complex structure consisting of a variety of ECM proteins and proteoglycans, including laminins, and is enriched with a variety of growth factors (Yurchenco et al., 2004). Since laminin families are thought to have an important role in the maintenance of BM integrity (Yurchenco et al., 2004), other components of the BM might not be properly organized or lost in the absence of laminin 1. Indeed, other previous and our present studies demonstrate that the loss of laminin chains causes thinner and disintegrated BMs, suggesting that the phenotypes in tab are caused by defects in overall BM signals mediated by laminin proteins and other ECM molecules. Regardless of this complexity, we found in the present study that the FAK-mediated pathway accounts for the phenotypes observed in the tab mutant, because the tab mutation abrogates the activation of FAK and an MO knockdown of medaka fak phenocopies neural tube defects in tab. The mosaic experiment confirmed the cell-autonomous function of FAK in neuroepithelial cells (Fig. 5D). ECM proteins exert their effects through a variety of cell-surface molecules, and integrins are most commonly used as signaling receptors (DeMali et al., 2003). Integrins are heterodimeric transmembrane receptors composed of α- and β-chains and more than 20 have so far been identified, of which α1β1, α3β1, α6β1, α7β1 and α6β4 are major laminin receptors (Desban et al., 2006; Nishiuchi et al., 2006). Although integrins activate a variety of tyrosine kinases and GTPases to induce multiple downstream signaling pathways, FAK is known to have a prominent role in integrin signaling (DeMali et al., 2003; Parsons, 2003).

We observed that activated FAK is localized on both the apical and basal sides of the neuroepithelial cells in wild-type embryos. The positive apical staining was unexpected, but a similar pattern for phosphorylated FAK was also observed in the neural tube of chick embryos (supplementary material Fig. S9), suggesting a conserved pattern of FAK activation in the vertebrate neural tube. Since both basally and apically localized pFAK becomes undetectable in tab mutants, their activation must depend on γ1-associated laminins. Although laminin 1 immunoreactivity was confined to the BM of the neural tube under our experimental conditions, we cannot rule out the possibility that other undetectable isoforms of laminins activate FAK on the apical side. Indeed, the apical region of the E10 murine cortex expresses Lama2, Lama4 and Lamb2 (Lathia et al., 2007), all of which are associated with Lamc1 and could thus account for the Lamc1-dependent apical activation of FAK.

Time-lapse imaging of living medaka embryos enabled us to monitor abnormal nuclear behavior in the mutant neuroepithelium before and after their division. A number of cells exhibit a reduced rate of nuclear migration toward the apical side without affecting the length of cell cycle, and prematurely undergo mitosis before they reach the surface region. Some mutant nuclei exhibit reversed migration toward the basal side before entering mitosis. This series of observations clearly demonstrates that mitosis and nuclear migration are decoupled in the tab mutant, suggesting the presence of a coupling mechanism under normal conditions. A similar situation has been reported for INM in the Pax6 mutant cortex in rat (Tamai et al., 2007). Furthermore, time-lapse imaging of wild-type and tab nuclei revealed an interesting feature of nuclear migration: wherever mitosis takes place (one exception in 25 tab mutants), the nuclei change their direction after having undergone mitosis. A mitotic event must therefore activate a specific signal that dictates the direction of the ensuing nuclear migration. These factors could regulate the direction of migration by switching motor molecules from dynein to kinesin (or vice versa), upon mitosis. Thus, using the medaka tab as a model system will provide a good starting point to further understand this putative switching mechanism.

The tab neural tube simultaneously exhibits abnormal INM and randomized mitotic orientation on the apical side, raising the question as to whether the misorientation of the cleavage plane affects INM. However, the two events are likely to be independent, because Konno and co-workers (Konno et al., 2008) have demonstrated that INM normally occurs in the LGN-mutant mice in which the orientation of normally planar divisions are randomized in the developing neocortex.

In tab neural tubes, the number of neurons is increased whereas the mitotic index is decreased at the neurogenic stage (15-somite stage) (Fig. 5B,C). Although we failed to examine the cell cycle exit rate because of technical difficulties, the result strongly suggests that neurogenesis in tab is accelerated at the expense of progenitor cells. This notion is consistent with previous reports in which impaired INM caused accelerated neurogenesis in the mouse cerebral cortex and zebrafish retina (Del Bene et al., 2008; Xie et al., 2007). Basally located mitosis might reduce the chance of daughter cells to be exposed to proliferative signals such as the Notch signal, which was reported to be high in the apical region of the zebrafish retina (Del Bene et al., 2008). Furthermore, Konno and colleagues recently demonstrated in the mammalian cortex that only planar division gives rise to apical progenitors which inherit both the apical and basal component of the mother cell (Konno et al., 2008). Our mosaic experiments provide a possible link between early defects in mitosis and accelerated neurogenesis at the later stage. Taken together, it is reasonable to conclude that both defective INM and mitotic orientation caused the observed accelerated neurogenesis in tab.

Previous studies implicated actin and the microtubule cytoskeleton in INM (Messier and Auclair, 1973; Messier and Auclair, 1974)and cleavage plane orientation (Siller and Doe, 2009; Toyoshima et al., 2007). Recently, Norden and co-workers (Norden et al., 2009) showed that INM in the zebrafish retina largely depends on the activity of actomyosin. However, in our present study, we put a special emphasis on studying the role of microtubules in medaka neurogenesis, partly because treatment with the actin-depolymerizing agent, cytochalasin B, was less effective in this system. Under our experimental conditions, treatments with high doses of cytochalasin B, which completely disrupted the actin network, induced mislocalized mitosis and misorientation of the mitotic cleavage plane in the medaka neural tube. However, this does not seem to reflect the tab mutant situation, because the actin network appeared normal in mutant neuroepithelial cells as indicated by staining of actin filaments in tab mutant neural tubes (supplementary material Fig. S10).

Dynein and microtubules are known to play evolutionarily conserved roles in spindle orientation and nuclear positioning (Siller and Doe, 2009). Furthermore, a recent zebrafish study revealed a crucial role for dynein and microtubules in INM during retinal neurogenesis (Del Bene et al., 2008). The zebrafish mok locus, the associated mutation of which causes severe INM defects in the retina, was found to encode the dynactin 1 subunit of the dynactin complex, a key regulatory complex during dynein motor activity. This led us to analyze the role of dynein and microtubules in INM and mitotic orientation in medaka and we found that the basal displacement of mitotic nuclei and randomized mitotic plane was induced when the dynactin complex was disrupted.

At present, little is known about a direct link between FAK-mediated signaling and the dynein-dynactin complex, although they both are involved in many common biological phenomena such as cell migration, proliferation and the establishment of cell morphology (Gomes et al., 2005). However, a recent report demonstrated that phosphorylated FAK is crucial for both neuronal migration and nuclear movement through the regulation of a small network of microtubules that encompass the nucleus (Xie et al., 2003). In more recent reports, phosphatidylinositol 3-kinase (PI3K), which is one of the downstream molecules of integrin signaling, was found to be essential for spindle orientation control by regulating accumulation of dynactin to the midcortex of a cell which astral microtubules contact (Toyoshima et al., 2007; Toyoshima and Nishida, 2007). Because FAK is also known to activate PI3K (Chen et al., 1996), FAK is likely to regulate mitotic machinery through the PI3K-mediated dynactin function, whereas in the case of INM, the activity of perinuclear dynein-dynactin can be a target of FAK signaling. Consistently, the velocity of nuclear migration decreases whereas the distribution pattern of stabilized and dynamic microtubules appears intact in tab neural tubes (supplementary material Fig. S11).

In support of the possible functional link between FAK-mediated signaling and the dynein-dynactin complex, co-injection experiments using suboptimal doses of dynamitin mRNA and fak MO demonstrated that they genetically interact for the regulation of INM and planar cell divisions. Furthermore, neural tubes of the fak morphants exhibit reduced dynactin 1 signal, which otherwise largely co-localizes with phosphorylated FAK in the apical and basal regions (supplementary material Fig. S8A,B). In tab neuroepithelial cells, the basal accumulation of dynactin appeared to slightly decrease (data not shown). However, overexpressed dynamitin did not alter the distribution pattern of phosphorylated FAK in the neural tube (supplementary material Fig. S8C), suggesting that the function of dynein-dynactin is downstream of FAK. It is still unknown how FAK regulates dynactin.

Since the overall phenotype following the co-injection resembles that obtained from a single injection of either dynamitin mRNA or fak MOs, the interaction between dynein and FAK appears to occur not only in the neural tube, but also in other developing organs. Further analysis, especially by biochemical approaches, will be needed to address how FAK-mediated signaling interacts with dynein-dynactin complexes to regulate a variety of biological processes, including INM and planar divisions in the developing neuroepithelium.

All studies of wild-type medaka were carried out using the d-rR strain of a closed colony. ENU-based mutagenesis was performed as described previously (Ishikawa, 1996). For mosaic analysis, medaka DsRed transgenic line was used.

Tab heterozygous fish were mated with wild-type HNI fish to generate F1 families. Homozygous mutant embryos were obtained from F1 crosses. Mapping was performed by using EST markers, BAC chromosome walking and using information in the medaka genome database.

Morpholinos (GeneTools) and capped RNAs synthesized using mMESSAGE mMACHINE (Ambion) were injected into 1-cell-stage drR embryos at a concentration of 150 μM for lamc1 MO1, 5′-TGTGTTGGTGATCTTACTTTTCTGC-3′; lamc1 MO2, 5′-CACAGTCCGCACAAAACTCTGCATC-3′, 100 or 300 μM for fak MO1, 5′-CCAGGTAGGAAGACGATGCCATACC-3′; fak MO2, 5′-GCTAACTGGCCATCATATTCTGAAA-3′; and fak 5mis MO, 5′-CCAcGTAgcAAGACcATGCgATCC, 0.05 or 0.4 mg/ml for dynamitin mRNA, and LacZ mRNA. For live imaging of nuclear migration and cell shapes, 0.3 mg/ml EGFP mRNA was injected at the 8- to 16-cell stage to label cells in a mosaic fashion.

For histological analysis, fixed embryos were embedded in Technovit 8100 (Heraeus Kulzer). For ultrastructural studies, embryos were fixed in 2% paraformaldehyde and 2% glutaraldehyde in 0.1 M sodium cacodylate buffer, postfixed in 1% OsO₄, stained with uranyl acetate and embedded in Epon (NEM). Sections were stained according to standard procedures and visualized with a JEM-1011 microscope (JEOL).

Whole-mount immunostaining of medaka embryos was performed as described previously (Koshida et al., 2005). The antibodies used were as follows; phospho-Histone H3 (Upstate, 1:200), laminin 1 (Sigma, 1:200), FAK [pY397] (BioSource International, 1:200), fibronectin (Sigma, 1:200), aPKCξ (Santa Cruz, 1:200), ZO-1 (Sigma, 1:200), γ-tubulin (Sigma, 1:1000) and dynactin p150 Glued (BD, 1:200). Alexa Fluor 488/555 goat anti-rabbit/mouse IgG (Molecular Probes) were used as secondary antibodies with DAPI and/or Rhodamine or Alexa Fluor 488-Phalloidin (Molecular Probes). For analysis of laminin 1 and γ-tubulin, embryos were permeabilized with acetone at −20°C for 5 minutes. For analysis of Hu C/D (BD, 1:500), BrdU (Serotec, 1:500) and PCNA (Sigma, 1:500), embryos were processed for cryosectioning (12 μm). To detect BrdU, cryosections were treated with 2N HCl for 1 hour. These specimens were observed by Zeiss LSM710 confocal microscope.

10 mM BrdU (Sigma) was injected into the yolk at the 10-somite stage. After incubation at 28°C for 2 hours, injected embryos were fixed.

Labeled medaka embryos were dechorionated with medaka hatching enzyme, immobilized with 3.5 mM heptanol and embedded in low melting point agarose (Sigma) filled with the heptanol solution. Optical z sections were collected at 2 μm steps at 3 minute intervals, starting at the 9-somite stage until the 12-somite stage with an Olympus FV500 confocal microscope at an ambient room temperature of 28°C. For measurement of the velocity of nuclear migration, the distance traveled by individual nuclei from when they started migration was measured with Imaris and ImageJ programs. For observation of cell shape, z sections were collected at 1 μm steps, and also using fixed embryos labeled with EGFP and phalloidin.

Protein extracts were prepared from medaka embryos at 5 d.p.f. as described previously (Henry et al., 2001). Anti-pFAK (1:2000) or anti-dynactin p150 (1:250) antibodies, and anti-rabbit/mouse HRP secondary antibodies (1:5000; Amersham) were used.

For mosaic analysis, cell transplantation experiments were performed as reported previously (Shimada et al., 2008) with the following modifications. A medaka transgenic line with ubiquitous expression of DsRed was used as a donor, and the donor embryos were injected with fak MO or 5mis MO. The transplanted embryos were fixed at either the 10- or 15-somite stage.

We thank Masato Kinoshita for providing a medaka DsRed line, K. Hatta, J. Wittbrodt, J. R. Martinez-Morales, Y. Oka, R. Kamiya, H. Yanagisawa and M. Kyokuwa for technical advice and helpful discussion, and Y. Ozawa for excellent fish care. This work was supported in part by Grants-in-Aid for Scientific Research Priority Area Genome Science and Scientific Research, Global COE Program from the Ministry of Education, Culture, Sports, Science and Technology of Japan. S. Tsuda was a recipient of a Fellowship of the Japan Society for the Promotion of Science for Junior Scientists.